Hello all,
Is anyone working on a way of using PD in a clustered environment? High-load objects could operate over a cluster. Ideally it would be nice if one could tell any object to run on the cluster, to speed up pix_processing or render more complex gem scenes.
I'm really thinking here of PD as a possible graphical interface to show the datapaths in a certain work process which may run on multiple machines and pass data back and forth. Rather than each object representing a machine, each object could represent a job on a machine. Well its an idea anyhow. (I'm thinking of prototyping something using SPREAD)
Anyone written, or interested in, a multicast netsend/netreceive for pd? For those not familiar with multicast this would allow the same data to be sent to almost any number of machines simultaneously. using only one multicast IP. (If they have a multicast enabled network)
Anyhow I think thats enough ranting for today.
Thanks Ben
This reply is a little late in coming, but hopefully is still relevant. I am a graduate student and I will be starting a class this semester in which I plan on pursuing pd clustering. But my interest is multi-channel audio, so my approach is quite different. But since the professor, Ken Perlin, is interested in the same areas as you, my approach might change.
Basically, the way I currently envision doing this is to intercept all messages sent to the pd core (there is a patch to pd that enables this) and send them to all pd instances running in the cluster. This would create exact copies of the patch on each computer in the cluster. Then I would need to create objects which facilitate control and interoperation of the individual instances of pd. But from what I little I know about the video stuff in pd, I don't think this approach would be very helpful, unless it was for multichannel video.
As for multicast, I haven't heard anything about using it with PD, but there is a project that was posted on this list a couple months ago where they were planning on dealing with the jitter in the netsend/netreceive objects so that their delay would be stable. I don't remember the details though.
.hc
On Thursday, Dec 19, 2002, at 19:42 Europe/Vienna, bbogart wrote:
Hello all,
Is anyone working on a way of using PD in a clustered environment? High-load objects could operate over a cluster. Ideally it would be nice if one could tell any object to run on the cluster, to speed up pix_processing or render more complex gem scenes.
I'm really thinking here of PD as a possible graphical interface to show the datapaths in a certain work process which may run on multiple machines and pass data back and forth. Rather than each object representing a machine, each object could represent a job on a machine. Well its an idea anyhow. (I'm thinking of prototyping something using SPREAD)
Anyone written, or interested in, a multicast netsend/netreceive for pd? For those not familiar with multicast this would allow the same data to be sent to almost any number of machines simultaneously. using only one multicast IP. (If they have a multicast enabled network)
Anyhow I think thats enough ranting for today.
Thanks Ben
PD-dev mailing list PD-dev@iem.kug.ac.at http://iem.kug.ac.at/cgi-bin/mailman/listinfo/pd-dev
-
bbogart -
Hans-Christoph Steiner